Institutional Herding, Positive Feedback Trading and Opening
Price Behavior in Taiwan
Chaoshin Chiao
Department of Finance, National Dong Hwa University, Hualien, Taiwan
Weifeng Hung†
Department of Finance, Da-Yeh University, Chang-Hua, Taiwan
Cheng F. Lee
Department of Finance, Rutgers University, Piscataway, New Jersey, USA
Preliminary draft 2007.3.24
Current version 2007.5.16
†
Corresponding author, E-mail address: wfhung@mail.dyu.edu.tw. Phone: 886-4-8511888, Ext. 3522.
Weifeng Hung thanks the financial support from National Science Council in Taiwan. The program
number is NSC95-2416-H-212-011.
Institutional Herding, Positive Feedback Trading and Opening
Price Behavior in Taiwan
Abstract
This paper investigates the cross-sectional relation between opening price behavior and the
institutional trading in the Taiwan stock market. As a result, inconsistent with the finding of
Lee, Lin, and Liu (1999) that Taiwanese institutions follow neither positive-feedback nor
negative-feedback trading strategies, we find that the institutional investors do herd, which is
mostly driven by positive feedback trading rather than price impact or institutional forecasting
ability. Moreover, the source of positive feedback trading comes from not only the returns
measured over the past trading day but also over the opening session. Finally, the institutional
positive feedback trading is more pronounced for small stocks than large ones.
JEL Classification: G10; G14; G15; G18
Keywords: Institutional Herding; Positive Feedback Trading; Opening price
1
1. Introduction
Herding is the tendency for investors of a particular group to buy or sell the same stocks over
the same time. Informational asymmetry may cause uninformed but rational speculators to
choose to trade in the same way as informed traders (Bikhchandani, Hirshleifer and Welch,
1992; and Banerjee, 1992). The recent studies reveal that institutions herd because they are
positive feedback (momentum) traders, buying stocks that have recently increased in value
and sell those that have recently declined, and their herding moves prices (Lakonishok,
Shleifer, and Vishny 1992; Grinblatt, Titman, and Wermers 1995; Nofsinger and Sias 1999;
and Wermers 1999).
Most of these studies have investigated the institutions herding using the U.S. data
(Grinblatt, Titman, and Wermers 1995; Nofsinger and Sias 1999; and Wermers 1999; Griffin,
Harris, and Topaloglu 2003).1 However, although these studies provide important evidence
on U.S. markets and institutional shareholders, the applicability of the findings to other
markets with different features is questionable.2 For example, for some fast emerging markets,
such as the Taiwan stock market, unique market structures and regulations may change the
well-documented institutional behavior and their impacts on stock prices. The purpose of this
paper aims to examine the influence of such market structures and regulations on the
institutional trading in the Taiwan stock market.
The Taiwan stock market is known as a fast globalizing and institutionalizing market.
Since the early 1980s, the Ministry of Finance of Taiwan has globalized its stock market,
widely dominated by individual investors (Harrison, 1994), in order to enhance its efficiency.
After two decades, its achievements have been recognized. According to, up to 31.3% of
dollar trading volume in the Taiwan stock market is attributable to trades by institutional
investors from 2001 to 2003. Contrasting this with the 70% share of all dollar trading volume
held by institutional investors on the NYSE in 1989 (Schwartz and Shapiro, 1992), and 3% in
1
Several studies propose that institutional herding is often accompanied by positive-feedback strategies.
Positive-feedback traders increase purchases in a particular security when it has recently performed well, and sell
when it has performed poorly. Although such trading practices can be interpreted as rational learning through
prices (Grossman and Stiglitz, 1976; Hellwig, 1980), many economists raise concerns that institutional trading
could be destabilizing and trigger ‘informational cascades’ that investors learn information by observing the
actions of others (Banerjee, 1992; Bikhchandani, Hirshleifer, and Welch, 1992). Moreover, the known presence
of feedback traders may prompt other investors to trade in a way that moves prices further away from their
efficient values (DeLong, et al., 1990). Froot, Scharfstein, and Stein (1992) and Hirshleifer, Subrahmanyam, and
Titman (1994) posit that investigative herding occurs when institutional investors’ information is positively
cross-sectionally correlated, possibly because they follow the same signals
2
Using data on foreign investors (or U.S. investors) in Korea as a single group, Choe, Kho, and Stulz (1999)
find evidence of herding.
1
Taiwan over the same year, we can see that institutional trading is still low but has rapidly
increased over time.
In addition, the stock trading in Taiwan is subject to regulatory price limits. The
government officials have imposed a daily price limit, 7%, both upward and downward, based
on the previous day's closing price of each stock traded in the TSEC. Any stock hitting its
price limit could still be traded as long as the transaction price was within the upper and lower
bounds. The purpose of the price limit is to prevent stocks from excessive volatility, to
counter overreaction, and to protect investors from potential daily losses. However, Kim and
Rhee (1997) supports that price limits prevent prices from efficiently reaching their
equilibrium level. According to Chang, McLeavey, and Rhee (1995), impacts of a price limit
tend to diminish as the return horizon extends.
Given the market structure and regulation absent in mostly developed markets, many
hypotheses held in mostly developed markets are worthwhile to be re-examined. In this paper,
the following questions are of interest: Does herding exist in Taiwan stock market? If so, what
factors cause institutional herding, do they follow steps for each others or they use some
certain common signals, such as past returns? Does institutional herding move price or they
have forecasting ability in predicting stock returns? We add to the literature by focusing on
three issues. First, we explore the cross-sectional relationship between lag, concurrent, and
future stock returns and institutional net buying trading to investigate the trading behavior of
institutions in Taiwan. Second, we are the first to use returns measured at the opening session
to proxy for intradily extreme price changes, and to investigate the behavior of institutional
trading after large opening price changes.
The paper documents the following results. First, there is a positive concurrent relation
between institutional trading and stock returns. Second, inconsistent with the finding of Lee,
Lin, and Liu (1999), we document that the positive relation between institutional trading and
stock returns is driven by institutional positive feedback trading neither by price impact
caused by institutional herding or by the forecasting ability of institutional trading. Third, the
institutional positive feedback trading comes from not only returns measured over the prior
trading day but also returns measured at opening session. Fourth, the opening price changes
have a nonlinear and asymmetry impact on institutional trading. It seems that the extreme
opening price changes have the negative impact on institutional trading. Finally, inconsistent
with Lakonishok, Shleifer, and Vishny (1992), the institutional positive feedback trading in
Taiwan stock market is more pronounced for small stocks than large stocks.
Our contributions beyond the previous literature can be primarily placed on the
2
examination in depth the institutional behaviors, given the unique microstructure and
regulation as well as the fast-grow nature of the Taiwan stock market. As Taiwan has
gradually opened its financial markets and institutional trading increasingly has gained
importance, Taiwan’s development may arouse the interests of policy makers of other
emerging markets. Taiwan’s experience can assist them in establishing effective policies to
promote the efficiency and fairness of price discovery.
The paper proceeds as follows. Section 2 describes the sample. Section 3 investigates the
relation between the institutional trading behavior and stock returns. The source of positive
feedback trading is discussed in Section 4. Section 5 analyses the relation between firm size
and institutional positive feedback trading. Section 6 provides regressions analyses on the
relationship between and institutional trading and past returns measured over the prior trading
day and over the opening session. Finally, Section 7 concludes this paper.
2. The data
The daily stock prices, market index returns (including dividends), and stock trading data,
such as the trading volumes and dollar volumes, of institutional investors are obtained from
Taiwan Economic Journal (TEJ). The database includes all currently and historically listed
common stocks on the Taiwan Security Exchange Commission. Due to the limited
availability of institutional investor’s trading data, we use a sample 1423 trading days, range
from December 26, 2000 to September 22, 2006. The daily stock prices consist of the opening,
closing, high, and low prices. The number of firms in our sample monotonically increases
over time. There are average 188 firms and 330 firms with complete information for each
trading day in 2001 and 2006.
Institutional investors in the Taiwan stock market are classified by the Taiwan Stock
Exchange Corporation into three major groups: foreign investors (FIS,thereafter), securities
investment trust companies (SITCS, thereafter), and securities dealers (SDS, thereafter).
Following Griffin, Harris, and Topaloglu (2003), for each individual stock, we calculate its
institutional trade imbalance as follows:
Institutio nal trade imbalance 
institutio nal buy volume - institutio nal sell volume
.
outstandin g shares
The positive (negative) institutional trade imbalance increases (decreases) the institutional
3
ownership for the stock. For each trading day, in order to avoid extremity due to illiquidity,
we exclude the stocks that ever hit the limit price during the trading day. Specifically, any
stock whose high (low) reaches its upper (lower) price limit will be excluded from the sample.
The stocks without institutional trading data are also excluded. The number of included stocks
increases from 113 on December 26, 2000 to 312 on September 22, 2006, and the average is
about 250.
3. The relationship between institutional trade imbalance and stock return
In this section, following Nofsinger and Sias’s (1999) definition of herding, we want to firstly
test whether the institutional herding exists in Taiwan stock market by focusing on the net-buy
and -sell activities of groups of investors who buy or sell the same stock over the same time.
If the herding exists, then second we attempt to show whether the institutional herding results
from either their following step for each other or their trading based on the same signals, such
as past stock returns. Specifically, we are going to distinguish whether the Taiwanese
institutional herding is in favor of the information cascade hypothesis or investigative herding
hypothesis. Finally, this paper defines the importance of herding by the extent to which the
positive relation occurs between institutional trade imbalances and stock returns measured
over the same time. That is, for a certain type of investors, its higher positive concurrent
relation between institutional trading and stock returns implies higher extent of herding.
We begin to explore the herding by investigating whether institutions have the same
trading direction upon the same stocks over the same period. We form two kinds of portfolios
which are separately sorted by institutional trade imbalances and stocks returns, and then
investigate the behaviors of institutional trading across trading days and portfolios. The
detailed procedure to form prior-returns-based portfolios is described as follows. The stocks
are divided into ten portfolios (deciles) based on the stock returns. The daily maximum price
ranges is 14% (for daily upper and lower limit price is ± 7% respectively). We divide daily
maximum price ranges into ten pieces of sub-ranges by 1.4%, and classify stocks according to
which sub-ranges the stock returns belongs to. For example, when stock returns is less than
-5.6% then the stock is classified into the loser (P01), if stock returns is between -5.6% and
-4.2% then the stock is assigned into portfolio 2, and when stock return is higher than 5.6%,
then the stock is categorized into the winner (P10). The rest of portfolios are defined similarly.
The first row of Panel A in Table 1 shows that, on average, the whole institutional trade
4
imbalance monotonically increases from -0.435% for the largest net-selling decile to 0.45%
for the largest net-buy decile. The F-statistic, under the hypothesis that the institutional trade
imbalance across the ten portfolios is equal, rejects the equality at the 1% level. Does this
mean that Taiwanese institutions herd because they trade by following steps for each other.
For instance, it is possible that the above result mainly comes from certain group(s) of
investors and their investment strategies.
[Insert Table 1 here]
In this paper, we document that there are three reasons why different types of
institutions tend to trade different stocks. First, Chiao and Shao (2006) indicate that the
buy-sell dollars trading imbalances are higher for FIs. However, in this paper, Panel A of
Figure 1 shows that the magnitudes of net-buy or net-selling activities are stronger for SITCs
than for FIs. Since the measure of institutional trade imbalances in this paper is scaled by the
number of outstanding shares, the stocks traded by FIs may have larger outstanding shares
than those stocks traded by SIT. Therefore, the inconsistency may simply reflect the
difference in their investment strategies. That is, the stocks traded by FIs tend to be lager
market capitalization (higher stock price and higher the numbers of outstanding shares) than
those stocks traded by SITCs.
[Insert Figure 1 here]
Second, Panel B of Figure 1 suggests that the concurrent relations between institutional
trade imbalance and stock returns are not systematic present across three types of institutions.
For example, for the FIs investors, it is a decreasing tendency in institutional net-buying
shares for winner portfolios from P09 to P10, and an increasing tendency in institutional
net-selling shares for loser portfolios from P02 to P01. In contrast, the institutional net-buying
shares for SIT investors experience a monotonic increasing pattern for winner portfolios and a
monotonic decreasing pattern for loser portfolios. Therefore, it is reasonable to conjecture that
the features of underlying stocks trading FI are very likely to be different from those stocks
trading SIT and SD investors, since they have different trading behaviors for either extreme
high or low stock returns.
Third, we compute, on a daily basis, the cross-sectional correlations of institutional trade
imbalances between FIs, SITCs, and SDs over the selected day (day 0), the previous day (day
5
-1), and the following day (day +1). Table 2 reports the time-series averages of these
cross-sectional correlations. The t-statistics are computed from the time-series standard errors
of these correlations. The results of Table 2 show, firstly, weak cross-correlations of
institutional trade imbalances between three types of institutional investors over day 0. For
instance, the time-series averages are only -1.6%, 1.38%, and 1.4%, significant though,
between FI and SDs, FIs and SITCs, and SDs and SITCs, respectively. Secondly, resembling
the cross-correlations and the cross-autocorrelations are quite small. Finally, unlike the
cross-correlations, the cross-autocorrelations, it is worth noting that the own-autocorrelations
are strikingly high. For instance, the correlations over days -1 and 0 are 35.99% 30.10%, and
15.42% for SITCs, FIs, and SDs respectively. The evidence suggests that the institutions in
Taiwan tend to persistently follow their own steps to trade rather than infer information based
on others types of institutional trading. Therefore, we argue that the institutional herding in
Taiwan may not be mainly driven by the hypothesis of informational cascades (Banerjee 1992;
Bikhchandani, Hirshleifer, and Welch 1992). However, why do they still have similar trading
directions for those stocks on the basis of prior performance? Do they herd by buying or
selling the same stocks during a short period of time?
3
[Insert Table 2 here]
Thus far, we have excluded the possibility that institutional herding may come from
trading by mimicking for each other. In the following, we tend to investigate the importance
of institutional herding and what factors drive the herding, by exploring the relation between
institutional trade imbalances and stock returns measured over days –1, 0, and +1. In
particular, as mentioned before, we regard the impact of herding on stock price as more
pronounced for a specific type of institution if there exists a stronger positive relation between
their trade imbalance and stock return over day 0. Table 3 reports the time-series average of
these cross-sectional correlations. The t-statistics are computed from the time-series standard
errors of these statistics. The results of Table 3 suggest that across the three types of
institutions, from high to low, the relations between institutional trade imbalances and returns
over day 0 are 25.56%, 14.02%, and 3.39% for SITCs, FIs, and SDs respectively. This
suggests that herding of SITCs impact the stock prices the most. According to our previous
3
Another theoretical foundation, investigative herding, may provide a suitable explanation for our results.
Institutional investors react to the same information leads them to trade the same stocks over the same time
(Froot, Scharfstein, and Stein 1992; Hirshleifer, Subrahmanyam, and Titman, 1994).
6
finding that the Taiwanese institutional herding mostly occur within a certain type of investor,
that is, Taiwanese institutions tend to follow their own step. Therefore, we would expect that
there will be a higher autocorrelation for those investors who herd the most. Indeed, the
results of Table 3 are in favor of the conjecture that the autocorrelation is highest for SITCs
about 36% compared to second of FIs with 30% and last of SDs with 15%.
It is possible that the stock price could change earlier than, later than, or within the day
on which herding happens. Given the high own-autocorrelations documented in Table 2, a
positive relation between institutional trade imbalances and stock returns measured over day 0
may arise if: (1) institutional investors follow positive-feedback trading, (2) institutional
herding generate impacts on stock prices, and (3) institutions are informed traders and their
trades can predict stock returns. For example, Griffin, Harris and Topaloglu (2003) find that
the positive cross-sectional daily relation between stock returns and institutional trading is due
to institutional positively following past intradaily stock returns rather than predictability and
price pressure by herding. Chakravarty (2001) and Sias, Starks, and Titman (2001) conclude
that a price change measured over the same period as institutional ownership changes is due
to the price impact of the institutional trading. Chen, Jegadeesh, and Wermers (2000) find that
stocks managers buy outperform stocks managers sell by 2% per year after controlling for
various characteristics. In the following sections, we will examine the above three
possibilities by investigating the institutional trading behavior during, before, and after and
herding day.
[Insert Table 3 here]
3.1. The price impact of institutional herding
To explicitly investigate the price impact by institutional herding, in the subsection we
examine the relation between the institutional trade imbalances and stock returns over day 0.
The time-series averages of institutional trade imbalances portfolios are reported in Panel A of
Table 1. The results clearly suggest an increasing pattern of stock return with respect to
institutional trade imbalance. The portfolio with the highest institutional net-buy (net-selling)
experiences returns of 1.413% (-0.902%) over day 0 and the difference between the extremes
(P10–P09) is 2.315%. All these statistics are significant at the 1% level. This result is
consistent with the hypothesis that institutional buy (sell) herding has a positive (negative)
impact on stock prices.
7
On the other hand, the results shown in Panel B of Table 1 suggest that the institutional
trade imbalances across portfolios exhibit a monotonic increasing tendency from P01 to P09.
However, a reverse relation is found between the extreme winner portfolios. For winner
portfolios, the return on P10 experiencing the average 0.141% of the outstanding shares by
institutional net-buy is significantly less than that on P09 with 0.149% of the outstanding
shares. This indicates that the positive relation between concurrent stock returns and
institutional trade imbalance seems to be weaker than that of portfolios based on institutional
trade imbalance. This suggests that for the extreme portfolios based on prior-performance, the
stocks with higher institutional net-buying activities do not necessarily imply lower (higher)
concurrent stock returns. This inconsistency of the price impact between portfolios sorted on
institutional trade imbalance and portfolios sorted on prior returns will be further examined in
the following sections.
3.2. The positive feedback trading by institutions
Grinblatt, Titman, and Wermers (1995), Choe, Kho, and Stulz (1999), Grinblatt and Keloharju
(2000), and Chui, Titman and Wei (2000) document that institutional investors are usually
positive feedback traders. Traders who increase purchases in a particular security when it has
recently performed well, and sell when it has performed poorly are considered positive
feedback (or momentum) traders (Jegadeesh and Titman, 1993; Grinblatt, Titman and
Wermers, 1995). Alternatively, traders who are net seller (net buyer) after high (low) returns
are considered to be negative feedback (or contrarian) traders (Lakonishok, Shleifer, and
Vishny, 1994; Nofsinger and Sias, 1999). We examine the returns over day –1 (the
pre-herding day) to investigate whether institutional investors follow positive feedback
trading strategies in Taiwan.
The pre-herding return for each institutional trade imbalance portfolios is reported in
Panel A of Table 1. The pre-herding returns for the two extreme net-buy portfolios (P10 and
P09) are and 0.991% and 0.455%, respectively, while those for the two extreme net-sell
portfolios (P01 and P02) are -0.703% and -0.354%, respectively. All of them are significant at
1%. This indicates that the institutional investors in Taiwan stock market exhibit the positive
feedback trading tendencies, and is consistent with the U.S. findings. Similarly, in Panel B of
Table 1 the direction of pre-herding day returns for winner and loser portfolios are consistent
with that of herding-day institutional trade imbalances. The pattern of the portfolios supports
a positive feedback institutional trading behavior.
8
3.3. The forecasting ability of institutional trading
Many papers have investigated the performances of institutional investors in different markets.
For example, Choe, Kho, and Stulz (2004), using Korean data, show that domestic investors
have an edge over foreign investors in trading domestic stocks. Grinblatt and Keloharju (2000)
argue that foreign institutions perform better than domestic institutions. Using daily data for
the 16 largest Finnish stocks, Grinblatt and Keloharju (2000) find that the foreigners and
domestic financial corporations perform better than the domestic individual investors. Kang
and Stulz (1997) find that foreign ownership in Japanese firms does not generate a
significantly positive impact on the stock returns. If the profitable trading by institutions does
exist in Taiwan, given the observed high performance (Seasholes, 2004), it is intuitively
credible that individual investors can gather the buy-sell information of institutional investors
and thereby make profits by following their moves.
As reported in Panel A of Table 1, the results of post-herding returns indicate that the
institutional herding is not irrational. The portfolios with the largest net-bought (P09, and P10)
experience statistically significant positive returns in the post-herding day (0.128% and
0.354%). The stocks with institutions net-sold (P01 and P02) continue to experience
statistically significantly negative returns (-0.282% and -0.139%). Thus far, we see that
institutions do know when to buy (sell), as these stocks continue to perform better (poorly)
from the herd-bought (herd-sold) day. Further Panel A suggests that the institutional trading is
persistent to the post-herding day. The two extreme net-sold portfolios continue to decrease
stocks ownership in the post-herding day (-0.152% of the numbers of outstanding shares for
P01, -0.054% of the numbers of outstanding shares for P02). The extreme net-bought
portfolios still continue to increase stocks ownership in the post-herding day (0.172% of the
numbers of outstanding shares for P01, 0.058% of the numbers of outstanding shares for
P02).
There are three possible explanations for our results. First, institutions may have private
information, which leads to correct trading decisions. Second, the persistent institutional
trading to the post-herding day may cause pressures on the stock price. Third, we could be
observing some sort of short-run return momentum behavior, especially because the
herding-day returns for these stocks are relatively large. Therefore, the alternative explanation
for the positive (negative) returns for the net-buy and net-sell stocks is that these firms are
experiencing stock return continuations independent of institutional behavior. Jegadeesh and
9
Titman (1993) and Chan, Jegadeesh, and Lakonishok (1996) identify that stock momentum
occurs in the U.S. stock market.
Panel B of Table 1 indicates first that the continuous post-herding returns for portfolio
sorted by stock returns are only two of ten significantly different from zero, they are P01 and
P10. Second, except that the extreme portfolios in both sides exhibit continuation in stock
return, the post-herding returns tend to be a reversal tendency. The current evidence suggests
that we can not explicitly conclude either the absence or the presence of the institutional
forecasting ability. There are four of ten portfolios (P01, P02, P09, and P10) in Panel A of
Table 1 exhibit significantly continuation in post-herding returns. However, in Panel B of
Table 1, except the two extreme portfolios (P01 and P10) having continuations in
post-herding stock returns, there are also two portfolios (P02 and P09) have reversal
post-herding returns. Therefore, in order to justify these confounding results between Panel A
and B, that is, whether the institutional forecasting ability is an artifact of a return
continuation behavior in Taiwan stock market, we will conduct alternative experiment in the
next sections.
3.4. Institutional trading after extreme opening price changes
The results in Section 3.1 and 3.3 can not justify the presence of both price impact causing by
institutional herding and institutional forecasting ability. Therefore, it is necessary to further
examine the existence of both effects. On the other hand, Griffin, Harris and Topaloglu (2003)
document that the positive cross-sectional daily relation between stock returns and
institutional trading is due to institutional positively following past intradaily stock returns
rather than predictability and price pressure by herding. Therefore, it is necessary to examine
the behavior of institutions within the trading day. We use the intradaily opening price
changes to proxy for the extreme price changes. The opening price changes are defined as:
Opening price change (OPC) 
Opening price on day t - Closing price on day t-1
.
Closing price on day t-1
In order to investigate the impact of institutional herding, we also define the intradaily
closing price changes as:
10
Closing price change (CPC) 
Closing price on day t - Opening price on day t
.
Closing price on day t-1
We attempt to observe the trading behavior of institutions following the extreme opening
price changes by forming the ten portfolios based on each stock’s opening price changes.
Similar to the procedure of returns portfolios, first, the stocks are divided into ten portfolios
(deciles) according to the OPC. Second, we divide daily maximum price ranges 14% into ten
pieces of sub-ranges by 1.4%, and classify stocks according to which sub-ranges the stock
returns belongs to (refer to footnote 2). For example, when stock returns is less than -5.6%
then the stock is classified into the loser (P01), if stock returns is between -5.6% and -4.2%
then the stock is assigned into portfolio 2, and when stock return is higher than 5.6%, then the
stock is categorized into the winner (P10). The rest of portfolios are defined similarly.
The Table 4 illustrates the following results. First, the price impact by institutional
herding seem not exist, because there is intraday reversal in stock returns. For example, the
stock returns after large opening price changes are reversal. Therefore, consistent with the
prior results, the higher institutional net-buy does not necessarily lead to higher stock returns.
Second, positive feedback traders are for FI and SITCs. Especially, recall Table 3, the first
row indicates that the highest correlation between institutional trade imbalances and prior
returns are 19.86%, 9.04%, and 3.89% for SITCs, FIs, and SDs respectively, suggesting that
SITCs have the highest extent of positive feedback trading tendencies and SDs have the
lowest tendency toward buying (selling) the past winner (loser) stocks. The evidence is
consistent with the conjecture that the higher extent of positive feedback trading lead to
higher correlation between institutional trade imbalances and returns measured over the same
time. Third, by post-herding returns, the evidence of the forecasting ability of institutions is
relative weak. Further, according to the Table 3, the third row indicates that the correlations
between institutional trade imbalances and stock returns measured over the following day are
relatively low. For SITCS, FIS, and SD institutions, the correlation coefficients are 6.21%,
4.85%, and 1.95% respectively significant at 10%. We reject the null hypothesis that there is
no relationship between institutional herding and the subsequent day’s returns. However, the
evidence that institutional forecasting ability is relatively weak.
Since we cannot recognize the behavior of institutions during the first trades of the day,
there is a problem that the institutional trade imbalances might also contain the trading which
occurs during the opening section. Therefore, using the intradaily data, we calculate the
institutional trade imbalances at the first session for each ten OPC portfolios from 2002/09/02
11
to 2006/05/30. The results of Table 5 are consistent with our argument. For example, except
the SDs investors, the institutional trade imbalance for SITs and FIs shows monotonically
increasing across the OPC portfolios from lowest to the highest.
In sum, the evidence shows that the positive relation between institutional trade
imbalances and stock returns is due to institutional positive feedback trading rather than price
impact and institutional forecasting ability, and mostly pronouncedly happen to SIT and FIs
investors. In contrast, minor evidence indicates that the SDs is negative feedback traders.
[Insert Table 4 here]
[Insert Table 5 here]
4. The source of institutional positive feedback trading
Last section indicates that the positive relation between institutional trade imbalances and
stock returns measured over the same day is mainly driven by institutional positive feedback
trading. However, in this section, we are going to examine what signals induce institutions to
act like a positive feedback trader, last day’s stock returns or/and today’s stock returns. That is,
we want to distinguish whether they follow the behaviors of interday and/or intraday stock
returns to trade.
In order to separately examine the effect of both factors, past returns and changes in
opening price, on the institutional trading behavior, we use a two pass independent sorting
procedure to allow variation in one variable while holding the other variable constant. Stocks
are first sorted into quintiles based on their changes in opening price. We then independently
sort the stocks into quintiles based on their past returns and form a five by five matrix of
portfolios independently sorted on opening price changes and past returns.
Table 6 reports the time-series average of the cross-sectional mean institutional trade
imbalances for each portfolio. Each column reports the institutional trade imbalances for
stocks that differ on opening price changes but experience similar past returns. The last
column reports F-statistics based on the null hypothesis that the institutional trade imbalances
does not differ across the past returns portfolios within each opening price changes quintile.
Analogous statistics are reported in the last row for the opening price changes portfolios
within each past returns quintile.
The results reported in Table 6 indicates that for FI and SITCs, both opening price
changes and past returns play a role in explaining the institutional trade imbalances. That is,
12
the F-statistics reported in the last column reveal that we reject the null hypotheses that the
changes in past returns exhibit equal subsequent institutional trade imbalances within each
opening price changes quintile. Similarly, F-statistics reported in the last row reveal that we
reject the null hypotheses that the changes in opening price exhibit equal subsequent
institutional trade imbalances within each past returns quintile. In contrast, consistent with
previous finding, the SDs have no tendency of following past returns or opening price
changes.
[Insert Table 6 here]
5. The positive feedback trading and market capitalization
Lakonishok, Shleifer, and Vishny (1992) document that pension fund feedback trading is
largely limited to smaller capitalization stocks. Wermers (1999) suggest that institutional
herding caused by information cascades more often appears within small capitalization groups.
On the other hand, Froot, Scharfstein, and Stein (1992) and Hirshleifer, Subrahmanyam, and
Titman (1994) indicate that the investigative herding are more likely to happen to those stocks
with larger market capitalization because the institutions have higher commonality in in
formation related to larger stocks. As we have found in previous sections, the institutional
herding in Taiwan stock market is mainly due to the investigative herding rather than
information cascades. Therefore, we might expect to see that the institutional herding is more
significant within the large capitalization stocks.
In this subsection, we want to test whether the positive relation between opening price
changes and institutional trade imbalances is more significant within the classification of
small market capitalization. Each day, we begin by sort the sample into three groups of
capitalization. Then within each capitalization group, we sort stocks into ten portfolios based
on its opening price changes. Table 7 reports the time-series average of the daily
cross-sectional mean level of institutional trade imbalances for each day for small and large
capitalizations. Inconsistent with Lakonishok, Shleifer, and Vishny (1992), we find that
institutional positive feedback trading is mainly limited to larger capitalization rather than
smaller capitalization, suggesting that the institutional herding in Taiwan stock market is due
to investigative herding.
13
[Insert Table 7 here]
6. The Regressions analysis
Given our hypothesis that the institutional trade imbalances are related to the opening price
changes and past returns. We adopt the Fama-MacBeth (1973) regressions to further examine
the relation between changes in opening price and institutional trade imbalances.
yi ,t   0,t  1,t LagRETi ,t   2,t NOPCi ,t  3,t POPCi ,t
(1)
  4,t SNOPCi ,t  5,t SPOPCi ,t  ei ,t
where independent variables LagRET is past returns, OPC is the percentage opening price
changes to previous closing price, SOPC is the square of OPC, NOPC equals OPC if OPC is
less than zero, POPC equals OPC if OPC is not less than zero. SNOPC is the square of NOPC,
SPOPC is the square of POPC. The dependent variable (y) is IN, FIS,SITCS, and SD
respectively.
Table 8 reports the cross-sectional mean coefficient estimates and the adjusted R2s. The
results of Table 8 suggest that the relation between institutional trade imbalances and past
returns and opening price changes are positive. The coefficients of past returns and OPC are
significant at 1%. However, particularly note that the relation between institutional trade
imbalances and opening price changes is nonlinearly negative. The results of Panel A show
that a significant nonlinear relation exists for both FI and SDs. For example, the average
coefficient of square opening price changes (SOPC) is -0.001% significant at 1% for FIs.
Furthermore, the previous results show that the opening price changes seem to have
negative impact on institutional trading when the opening price changes are in extreme level.
Especially, the extreme upper OPC has an obvious impact on institutional trading. Therefore,
we attempt to further test whether the asymmetry relation between institutional trading and
OPC in level and direction for upper and lower stock returns exist. Panel B of Table 8
indicates that consistent with the findings in previous sections, for three types of institutions,
the extreme positive opening price changes have negative impact on institutional trading
behaviors at 1% significance. Thus, we conclude that the OPC has nonlinear impact on
institutional trading.
[Insert Table 8 here]
7. Concluding remarks
14
Following Nofsinger and Sias (1999), we measure the extent of herding by investigating the
cross-sectional concurrent relation between stock returns and the institutional trade
imbalances (buy volume minus sell volume) in Taiwan stock market. We find that the
institutional herding does exist in Taiwan stock market, and consistent with the hypothesis of
investigative herding, the institutional herding results from their positive feedback trading.
Little evidence is found that they infer the information from others trades. For further
justification the source of herding, we use opening price changes to proxy for the extreme
intradily price changes, and examine the institutional behaviors after the extreme opening
price change.
We find that the Taiwanese institutional herding, especially for foreign investors and
securities investment trust companies, is mostly driven by institutional positive feedback
trading rather than price impact or institutional forecasting ability. Moreover, the source of
institutional positive feedback trading not only comes from the returns measured over the past
trading day but also come from returns measured over the opening session. That is, in addition
to the past stock returns, we find that the opening price changes also play a crucial role in
explaining the institutional trading behaviors. In contrast to the positive influence of past
returns on institutional trading, the extreme concurrent opening price has negative impacts on
institutional trading. Moreover, inconsistent with Lakonishok, Shleifer, and Vishny (1992), we
find that the institutional positive feedback trading is mainly present in larger market
capitalization stocks. Finally, we find little evidence that the securities dealers herd and follow
the positive feedback trading strategy.
15
Reference
Banerjee A., 1992, “A simple model of herd behavior,” Quarterly Journal of Economics 107,
797-817
Bikhchandani, S., D. Hirshleifer, and I. Welch, 1992, “A theory of fads, fashion, custom, and
cultural change as informational cascades,” Journal of Political Economy 100, 992-1026.
Chan, L.K.C. and Josef Lakonishok, 1995, “The Behavior of stock prices around institutional
trades, ”Journal of Finance 50,1147-1174.
Chang, R.P., D.W. McLeavey, S.G. and Rhee 1995, “Short-term abnormal returns of the
contrarian strategy in the Japanese Stock Market,” Journal of Business, Finance, and
Accounting 22, 1035-1048.
Chiao, C. and K.I. Lin (2004), “The informative content of the net buy information of
institutional investors: evidence from the Taiwan stock market”, Review of Pacific Basin
Financial Markets and Policies, 7, 259-288.
Chiao, C., K. Hung, and C .F. Lee (2004), “The price adjustment and lead-lag relations
between stock returns: microstructure evidence from the Taiwan stock market”, Journal of
Empirical Finance, forthcoming.
Choe, H., B.C. Kho, and R. M. Stulz. 1999. Do foreign investor destabilize stock markets?
The Korean experience in 1997. Journal of Financial Economics 54, 227-264.
Dahlguist, M. and G. Robersson, 2001, “Direct foreign ownership, institutional investors, and
firm characteristics,” Journal of Financial Economics 59, 413-440.
DeLong, J. B., A. Shleifer, L. Summers, and R. Waldmann, 1990, “Positive feedback
investment strategies and destabilizing rational speculation,” Journal of Finance 45, 379-395.
Falkenstein, E.G., 1996, “Preferences for stock characteristics as revealed by mutual fund
portfolio holding,” Journal of Finance 51, 111-135.
Froot, K.A., D.S. Scharfstein, and J.C. Stein, 1992, “Herd on the street: Informational
inefficiencies in a market with short-term speculation,” Journal of Finance 47, 1461-1484.
Grinblatt, M. and M. Keloharju. 2000, The investment behavior and performance of various
investor types: A study of Finland’s unique data set. Journal of Financial Economics 55:
43-67.
Grossman, S. and J. Stiglitz, 1976, “Information and competitive price systems,” American
Economic Review 66, 246-253.
Harrison, M., 1994, Asia-Pacific Securities Markets, 2nd Ed. Hong Kong: Longman.
Hasbrouck, J., 1991, Measuring the information content of stock trades, Journal of Finance
46, 179-207.
16
Hellwig, M., 1980, “On the aggregation of information in competitive markets,” Journal of
Economic Theory 22, 477-498.
Hirshleifer, D., A. Subrahmanyam, and S. Titman, 1994, “Security analysis and trading
patterns when some investors receive information before others,” Journal of Finance 49,
1665-1698.
Kang, J.-K., Stulz,R.M., 1997, “Why is there a home bias? an analysis of foreign portfolio
equity ownership in Japan,” Journal of Political Economics 46, 3-28.
Kim, K. and S.G. Rhee, 1997, “Price limit performance: evidence from the Tokyo Stock
Exchange,” Journal of Finance 52, 885-899.
Lakonishok, J., A. Shleifer, and R.W. Vishny, 1992, “The impact of institutional trading on
stock prices,” Journal of Financial Economics 32, 23-43.
Lee, Y.T., J.C. Lin, and Y.J. Liu, 1999,“Trading patterns of big versus small players in an
emerging market: An empirical analysis,”Journal of Banking &Finance 23, 701-725.
Lewis, K.K., 1999, “Trying to explain home bias in equities and consumption,” Journal of
Financial Literature 37, 571-608.
Merton, R.C., 1987, “A simple model of capital market equilibrium with incomplete
information,” Journal of Finance 42, 483-510.
Nofsinger, J. and R.W. Sias, 1999, “Herding and feedback trading by institutional and
individual investors”, Journal of Finance 54, 2263-2295.
Scharfstein, D.S., and J.C. Stein, 1990, “Herd behavior and investment,” American Economic
Review 80, 465-479.
Schwartz, R.A., and J. Shapiro, 1992, “The challenge of institutionalization for the equity
market, recent development in finance,” in Anthony Saunders, ed., Recent Developments in
Finance, (Business One Irwin).
Sias, R., L. Starks, and S. Titman, 2001, “The price impact of institutional Trading,” working
paper, Washington State University and The University of Texas.
Sias, R., 2002, “Institutional herding”, Review of Financial Studies, forthcoming.
Titman, S., K.C. Wei, and T. Yamada, 2004, “Stock holding preference and performance
by different types of investors,” Proceedings of The 12th conference on the Theories and
Practice of Securities and Financial Markets, Kaohsiung.
Wermers, R., 1999, “Mutual fund herding and the impact on stock price,” Journal of Finance
54, 581-622.
Yu, C.H., 2002, “The role of institutional ownership and herding in diversification strategy,”
Journal of Risk Management 42, 1-29.
17
18
Figure 1. Portfolios sorted by daily institutional trade imbalances and stock returns
In Panel A, for each day from December 26, 2000 to September 22, 2006, totally 1423 trading days,
the all stocks in samples are ranked by their daily institutional trade imbalances and classified into
deciles portfolios. For each stock, the institutional trade imbalance is the difference between the
institutional buy and sell volumes for that day scaled by the total number of outstanding shares. In
Panel B, ten portfolios are sorted by stock returns. The maximum price ranges 14% are divided into
ten pieces of sub-intervals. The ten sub-intervals are between -5.6%, -4.2%, -2.8%, -1.4%, 0%, 1.4%,
2.8%, 4.2%, and 5.6%. Then each stock is classified into portfolio according to its stock returns
belonging to which price intervals. For example, for stock returns greater than 5.6% are categorized as
winner portfolio (P10), the stocks with returns between 5.6% and 4.2% are categorized into P09, and
so on. The time-series averages of institutional trade imbalances for each portfolio are reported. The
institutional trade imbalances and stock returns are expressed in percentage per day. FIs, SDs, and
SITCs are denoted as foreign investors, securities dealers, and securities investment trust companies,
respectively.
Panel A: Portfolios based on institutional trade imbalance
0.300
FIs
SITCs
SDs
Percentage of Institutional imbalances
0.200
0.100
0.000
P01
P02
P03
P04
P05
P06
P07
P08
P09
P10
Portflios sorted by institutional imbalances
-0.100
-0.200
-0.300
Panel B: Portfolios based on stock return
0.060
FIs
SDs
SITCs
0.100
Percentage of Institutional trade
imbalance (FIs and SDs)
0.040
0.060
0.040
0.020
0.020
0.000
0.000
P01 P02 P03 P04 P05 P06 P07 P08 P09 P10
Portflios sorted by reutrns
-0.020
-0.020
-0.040
-0.060
-0.080
-0.040
-0.100
-0.060
-0.120
19
Percentage of Institutional trade
imbalance (SITCs)
0.080
Table 1. Institutional trade imbalances and price behaviors on portfolios separately sorted by daily institutional trade imbalance and
stock return
This table reports the average institutional trade imbalances and price behavior in percentage on portfolios separately sorted by daily institutional trade
imbalance in and stock return on deciles are reported in Panels A and B, respectively. For each stock, the institutional trade imbalance is the difference
between the institutional buy and sell volumes for that day scaled by the total number of outstanding shares, while price behavior is measured by the return
over the selected day. On each day (day 0) from December 26, 2000 to September 22, 2006, totally 1423 trading days, all stocks in our sample are ranked by
their daily institutional trade imbalance and classified into 10 portfolios. In Panel B, the maximum price ranges 14% are divided into ten sub-intervals. The
break points are -5.6%, -4.2%, -2.8%, -1.4%, 0%, 1.4%, 2.8%, 4.2%, and 5.6%. Then each stock is then classified into portfolios according to where its stock
return locates. For example, for a stock return greater than 5.6%, the stock is categorized into the winner portfolio (P10); a stock with a return between 5.6%
and 4.2% is categorized into P09, and so on. The time-series averages of the institutional trade imbalances, and lead, contemporaneous, and lagged stock
returns are reported. ***, **, and * denote level of significance at 1%, 5%, and 10% respectively.
Panel A: Deciles based on institutional trade imbalance
P01
P02
P03
P04
P05
Institutional trade imbalance (%)
Day 0
-0.435*** -0.118***
-0.047***
-0.017***
-0.004***
0.004***
0.018***
0.048***
0.120***
Price behavior (%)
Day -1
-0.703***
Day 0
-0.902***
Day +1
-0.282***
-0.219***
-0.275***
-0.071*
-0.119***
-0.142***
-0.039
-0.064*
-0.105***
0.019
-0.005
-0.056
-0.027
0.093***
0.038
-0.008
0.225***
0.272***
0.056
0.455***
0.618***
0.128***
-0.354***
-0.479***
-0.139***
P06
P07
P08
P09
P10
P10-P01
F-stat.
0.450***
0.885***
6355***
0.991***
1.413***
0.354***
1.693***
2.315***
0.635***
142***
255***
18***
P10-P01
F-stat.
0.287***
439***
Panel B: Deciles s based on stock return
P01
P02
P03
P04
P05
Institutional trade imbalance (%)
Day 0
-0.146*** -0.137***
-0.084***
-0.047***
-0.012***
0.019***
0.063***
0.110***
0.149***
0.141***
Price behavior (%)
Day -1
-0.537***
Day 0
-6.442***
Day +1
-0.321***
0.297***
-3.371***
0.085*
0.291***
-1.989***
0.024
0.068**
-0.482***
-0.038
-0.124***
0.759***
0.039
-0.098**
2.010***
-0.039
0.082*
3.422***
-0.169***
0.380***
4.818***
-0.324***
0.662***
6.511***
0.426***
-0.139**
-4.795***
0.246***
P06
P07
P08
P09
P10
1.199***
38***
***
12.953
722633***
***
0.747
20***
Table 2. Average cross-sectional correlation between institutional trade imbalances
For each day from December 26, 2000 to September 22, 2006, totally 1423 trading days, we estimate
the cross-sectional correlations. Panels A to C report the time-series mean correlation in percentage
between institutions over the previous, same, and following trading day. The correlation coefficients
are presented in percentage. FIs, SDs, and SITCs are denoted as foreign investors, securities dealers,
and securities investment trust companies, respectively. ***, **, and * denote level of significance at
1%, 5%, and 10% respectively.
Day -1
Day 0
Day +1
Panel A: FIs
FIs
SDs
SITCs
***
30.10
-0.26
2.96***
-1.16***
1.38***
30.16***
-0.09
2.48***
Panel B: SDs
FIs
SDs
SITCs
-0.08
15.42***
3.12***
-1.16***
1.40***
-0.41
15.45***
0.88**
Panel C: SITCs
FIs
SDs
SITCs
2.34***
0.91**
35.99***
1.38***
1.40***
-
2.98***
3.23***
36.09***
Table 3. Average cross-sectional correlation between institutional trade imbalances and
stock returns
For each day from December 26, 2000 to September 22, 2006, totally 1423 trading days, we estimate
the cross-sectional correlations between institutional trade imbalances and returns measured over the
previous, same, and following trading day for all stocks in samples. The correlation coefficients are
presented in percentage. FIs, SDs, and SITCs are denoted as foreign investors, securities dealers, and
securities investment trust companies, respectively. ***, **, and * denote level of significance at 1%,
5%, and 10% respectively.
Returns over day -1
Returns over day 0
Returns over day +1
OPC
FIs
9.04***
14.02***
4.85***
6.56***
SDs
3.89**
3.39**
1.95*
1.00*
1
SITCs
19.86***
25.56***
6.21***
10.21***
Table 4. Portfolios sorted by opening price changes
On each day from December 26, 2000 to September 22, 2006, totally 1423 trading days, the all stocks in TSE are ranked by their daily opening price changes
and classified into deciles portfolios. The maximum price ranges 14% are divided into ten pieces of sub-intervals. The ten sub-intervals are between -5.6%,
-4.2%, -2.8%, -1.4%, 0%, 1.4%, 2.8%, 4.2%, and 5.6%. Then each stock is classified into portfolio according to its ratio of opening price changes belonging
to which price intervals. For example, for stock’s OPC greater than 5.6% are categorized as winner portfolio (P10), the stocks with returns between 5.6% and
4.2% are categorized into P09, and so on. The time-series averages of institutional trade imbalances are reported. The institutional trade imbalances and stock
returns are expressed in percentage per day. FIs, SDs, and SITCs are denoted as foreign investors, securities dealers, and securities investment trust
companies, respectively. ***, **, and * denote level of significance at 1%, 5%, and 10% respectively.
P01
Past returns
Returns
Following returns
OPC
CPC
FIs
SDs
SITCs
P02
***
-2.427
-2.680***
-0.149
-6.431***
3.973***
-0.048***
-0.006
-0.060***
P03
***
-1.889
-2.388***
0.267**
-4.839***
2.345***
-0.033***
-0.004
-0.074***
P04
***
-1.195
-1.773***
-0.028
-3.353***
1.561***
-0.021***
-0.003
-0.050***
P05
***
-0.466
-1.284***
-0.011
-1.918***
0.642***
-0.011***
-0.002***
-0.027***
P06
**
-0.080
-0.376***
-0.023
-0.247***
-0.132***
0.001
-0.001***
-0.012***
P07
***
0.171
0.482***
-0.036
0.738***
-0.252***
0.012***
0.000
0.007***
P08
***
0.570
1.326***
-0.028
1.903***
-0.553***
0.020***
0.000
0.032***
P09
***
1.395
2.406***
-0.058
3.339***
-0.876***
0.020***
-0.001
0.044***
P10
***
2.418
3.081***
0.228**
4.818***
-1.607***
0.033***
-0.011***
0.060***
F-stat.
***
3.217
3.209***
-0.134
6.366***
-2.928***
0.016*
-0.007
0.068***
396.52***
985.23***
2.70**
252841***
708.15***
28.57***
2.10**
83.26***
Table 5. Portfolios sorted by opening price changes from 20020902 to 20060530: Imbalances at open session, during the day, and after
open session
On each day from September 2, 2002 to May 30, 2006, totally 929 trading days, the all stocks in TSE are ranked by their daily opening price changes and
classified into deciles portfolios. The maximum price ranges 14% are divided into ten pieces of sub-intervals. The ten sub-intervals are between -5.6%, -4.2%,
-2.8%, -1.4%, 0%, 1.4%, 2.8%, 4.2%, and 5.6%. Then each stock is classified into portfolio according to its ratio of opening price changes belonging to
which price intervals. For example, for stock’s OPC greater than 5.6% are categorized as winner portfolio (P10), the stocks with returns between 5.6% and
4.2% are categorized into P09, and so on. The time-series averages of institutional trade imbalances are reported. The ratios of imbalances traded at open
session to the imbalances traded during the day are expressed in parentheses. The institutional trade imbalances and stock returns are expressed in percentage
per day. FIs, SDs, and SITCs are denoted as foreign investors, securities dealers, and securities investment trust companies, respectively. ***, **, and *
denote level of significance at 1%, 5%, and 10% respectively.
P01
Imbalances at
open session
FIs
SDs
SITCs
Imbalances
during the day
FIs
P03
P04
P05
P06
P07
P08
-0.0018** -0.0015*** -0.0005*
-0.0002
0.0000 0.0005*** 0.0010***
***
***
0.0001 -0.0001
0.0000 -0.0001
-0.0001*** -0.0002*** -0.0006***
**
*
***
***
-0.0027
-0.0007 -0.0006
-0.0002
-0.0001 0.0006*** 0.0029***
P09
P10
F-stat.
-0.0006
-0.0070
-0.0009 -0.0022***
0.0071*** 0.0183***
2.20*
6.49***
17.80***
-0.0623*** -0.0362*** -0.0198*** -0.0096***
(8.06)
(5.05)
(7.40)
(5.31)
0.0017*
(-14.07)
0.0130***
(0.01)
0.0221***
(2.17)
0.0177***
(5.86)
0.0322***
(-1.78)
0.0119
(-59.20)
17.72***
-0.0050* -0.0023*** -0.0030***
(-1.66)
(5.94)
(1.31)
-0.0019
(6.54)
-0.0002
(51.49)
0.0019**
(-10.99)
-0.0017
(35.36)
-0.0081
(10.93)
-0.0026
(86.30)
1.00
SITCs
-0.0432*** -0.0458*** -0.0343*** -0.0213*** -0.0111***
(16.71)
(5.84)
(2.14)
(2.69)
(2.12)
0.0047***
(-1.51)
0.0261***
(2.25)
0.0386***
(7.47)
0.0536***
(13.17)
0.0666***
(27.45)
41.78***
FIs
SDs
SITCs
-0.0573*** -0.0344*** -0.0184*** -0.0091***
0.0019**
*
***
***
-0.0049
-0.0051 -0.0022
-0.0029
-0.0018
-0.0360*** -0.0432*** -0.0336*** -0.0207*** -0.0108***
0.0130***
-0.0001
0.0048***
0.0216***
0.0021**
0.0255***
0.0166***
-0.0011
0.0357***
0.0328***
-0.0072
0.0465***
0.0189*
-0.0003
0.0483***
18.73***
1.04
39.29***
SDs
Imbalances after
the open session
-0.0050**
0.0009
-0.0072
P02
-0.0040
(-21.98)
1
Table 6. Portfolios independently sorted on opening price changes and past returns
On each day from December 26, 2000 to September 22, 2006, totally 1423 trading days, we construct
a five by five matrix. First, the all samples are classified into five OPC portfolios. The maximum price
ranges 14% are divided into five pieces of sub-intervals. The five sub-intervals are between -4.2%,
-1.4%, 1.4%, and 4.2%. Then each stock is classified into five portfolios according to its ratio of
opening price changes belonging to which price intervals. For example, for stock’s OPC greater than
4.2% are categorized as highest OPC group (P10), the stocks with returns between 4.2% and 1.4% are
categorized into P04, and so on. Then the past return portfolios are independently constructed as the
same way. FIs, SDs, and SITCs are denoted as foreign investors, securities dealers, and securities
investment trust companies, respectively. The time-series averages of FIs, SDs, and SITCs imbalances
are reported in Panel A, B, and C respectively and are expressed in percentage per day. ***, **, and *
denote level of significance at 1%, 5%, and 10% respectively.
Low
OPC
Low
P02
P03
P04
High
8.95***
F-stat.
OPC
-0.060***
-0.043***
-0.023***
-0.004
-0.020
Low
P02
P03
P04
High
-0.008
-0.009***
-0.006***
-0.012***
-0.013
F-stat.
OPC
Low
P02
P03
P04
High
F-stat.
H-L
-0.112***
-0.094***
-0.076***
-0.022***
-0.002
14.52***
Low
P02
P03
P04
High
Past returns
P03
P04
High
F-Stat.
Panel A: FIs imbalances
-0.031***
-0.037**
-0.007
***
-0.023
-0.009***
0.000
-0.008***
0.005***
0.023***
0.014***
0.017***
0.031***
**
0.011
0.020
0.072***
-0.019*
0.010
0.041***
0.042***
0.015*
2.29*
18.26***
73.60***
14.69***
6.24***
19.11***
24.19***
7.91***
Panel B: SDs imbalances
-0.007**
-0.002
-0.009
-0.007***
-0.001
-0.002
-0.002***
-0.001***
0.002***
-0.003
0.001
0.003*
-0.019**
-0.011***
-0.003
0.004
0.004
0.010***
0.006***
0.001
5.51***
0.53
F-stat.
OPC
P02
20.36***
5.42***
1.59
1.18
Panel C: SITCs imbalances
-0.048***
-0.043***
-0.045**
***
***
-0.050
-0.021
-0.001
-0.029***
-0.008***
0.024***
-0.009***
0.024***
0.061***
0.004
0.033***
0.059***
0.009
0.024***
0.064***
0.095***
0.099***
16.41***
Low
-2.852***
-1.402***
0.184***
1.765***
3.032***
P02
-2.612***
-1.442***
0.030
1.379***
3.404***
305.54***
5.884***
590.35***
6.016***
35.52***
Panel D: Returns
P03
P04
-2.874*** -1.086***
-1.394*** -1.598***
-0.188*** -0.344***
1.347***
1.608***
***
3.271
3.000***
927.84***
5.546***
554.88***
6.145***
39.42***
Winner
-1.086***
-1.598***
-0.344***
1.608***
3.000***
364.32***
4.086***
0.77
3.58*
20.28***
8.41***
1.81
7.77***
49.71***
276.64***
100.13***
9.52***
20.79***
F-Stat.
11.10***
0.92
19.31***
5.78***
1.35
W-L
1.766***
-0.196
-0.528
-0.157
-0.032
Table 7. Portfolios sorted by opening price changes for sub-samples with different market capitalizations
On each day from December 26, 2000 to September 22, 2006, totally 1423 trading days, the all stocks in samples are, first, separated into three sub-samples
according to the stock’s market capitalization. Then within the each market capitalization sub-samples, we divide stocks into ten groups according to the
stock’s percentage of opening price changes to previous closing price (OPC). The maximum price ranges 14% are divided into ten pieces of sub-intervals.
The ten sub-intervals are between -5.6%, -4.2%, -2.8%, -1.4%, 0%, 1.4%, 2.8%, 4.2%, and 5.6%. Then each stock is classified into portfolio according to its
OPC belonging to which price intervals. For example, for OPC greater than 5.6% are categorized as winner portfolio (P10), the stocks with OPC between
5.6% and 4.2% are categorized into P09, and so on. The time-series averages of lead, contemporaneous, and lagged stock returns are reported and the
time-series averages of concurrent institutional trade imbalances, OPC, and CPC are also reported. The institutional trade imbalances and stock returns are
expressed in percentage per day. FIs, SDs, and SITCs are denoted as foreign investors, securities dealers, and securities investment trust companies,
respectively.
P01
P02
P03
P04
***
***
***
***
P05
P06
P07
P08
P09
P10
F-stat.
0.140***
0.449***
-0.015
0.748***
0.004***
0.000
0.002
0.494***
1.121***
-0.055
1.920***
0.005***
-0.002**
0.014***
1.386***
2.145***
-0.044
3.353***
-0.003
-0.002
0.024***
2.635***
2.669***
-0.202
4.806***
0.011
-0.010*
0.026***
3.394***
2.920***
-0.342**
6.391***
-0.007
-0.002
0.021*
291.29***
524.42***
3.62***
167797***
1.87
0.80
16.54***
0.579***
1.553***
-0.009
1.874***
0.051***
0.008***
0.039***
0.960***
2.761***
0.154
3.303***
0.060***
0.007*
0.066***
1.789***
3.734***
0.513***
4.804***
0.065***
-0.003
0.081***
2.849***
3.754***
0.624***
6.319***
0.120***
-0.014
0.135***
152.10***
712.18***
5.96***
86301.80***
69.52***
9.97***
80.87***
Panel A: Small firms
Past returns
Returns
Following returns
OPC
FIs
SDs
SITCs
-2.483
-2.305***
-0.070
-6.428***
-0.015
-0.003
-0.028***
-1.584
-2.023***
0.420***
-4.854***
0.001
0.000
-0.039***
-1.020
-1.554***
0.063
-3.382***
-0.003
-0.002
-0.015***
-0.378
-1.150***
0.050
-1.941***
0.000
-0.002
-0.013***
-0.102***
-0.319***
0.018
-0.257***
0.001
-0.001**
-0.007***
Panel B: Large firms
Past returns
Returns
Following returns
OPC
FIs
SDs
SITCs
-2.858***
-3.688***
-0.415
-6.376***
-0.105***
-0.017***
-0.118***
-2.663***
-3.553***
-0.144
-4.831***
-0.117***
-0.022***
-0.130***
-1.516***
-2.571***
-0.260**
-3.323***
-0.067***
-0.009**
-0.060***
-0.687***
-1.587***
-0.176**
-1.885***
-0.036***
-0.006***
-0.042***
-0.073***
-0.419***
-0.037
-0.241***
0.001
-0.002***
-0.015***
0.198***
0.542***
-0.044
0.722***
0.022***
0.002***
0.009***
Table 8. Daily cross-sectional regressions
From December 26, 2000 to September 22, 2006, totally 1423 trading days, each day we run the
following cross-sectional regressions. The dependent variables in the regressions are trading
imbalances by whole institutional investors, foreign investors (FIs), securities investment trust
companies (SITCs), securities dealers (SDs). The independent variables are past returns (LagRET), the
percentage opening price changes to previous closing price (OPC), the square of OPC (SOPC),
negative OPC (NOPC), positive OPC (POPC), the square of negative OPC (SNOPC), the square of
positive OPC (SPOPC). The average time-series coefficients are reported in the left hand side of the
table and the t-statistics are reported in the right hand side of the table. All variables are expressed in
percent per day. ***, **, and * denote level of significance at 1%, 5%, and 10% respectively.
Panel A: yi ,t  0,t  1,t LagRETi ,t  2,t OPCi ,t  3,t SOPCi ,t  ei ,t
Intercept
LagRET
OPC
SOPC
Adj-R2
FIs
Mean
SDs
SITCs
FIs
t-ratio
SDs
SITCs
0.005
0.005
0.008
-0.001
3.746
-0.001
0.001
0.001
0.000
2.695
-0.007
0.012
0.012
0.000
6.075
8.49
27.14
15.99
-4.65
-2.47
14.00
2.43
-3.55
-7.82
41.64
21.65
-0.40
Panel B: yi ,t  0,t  1,t LagRETi ,t  2,t NOPCi ,t  3,t POPCi ,t  4,t SNOPCi ,t  5,t SPOPCi ,t  ei ,t
Intercept
LagRET
NOPC
POPC
SNOPC
SPOPC
Adj-R2
FIs
Mean
SDs
SITCs
FIs
t-ratio
SDs
SITCs
0.004
0.006
0.009
0.012
0.001
-0.003
5.078
-0.001
0.001
0.002
0.003
0.001
-0.001
3.928
-0.010
0.012
0.010
0.023
0.002
-0.003
7.419
6.76
28.08
4.78
10.24
0.26
-6.20
-3.53
13.90
2.18
3.61
1.46
-2.74
-9.72
41.59
3.57
13.81
1.15
-4.60